Unfortunately the default permissions on NT are typically wide open, e.g. anyone can write to system directory and many registry keys. There was a time when Unix was like this, maybe Win2000 will fix this.
Re:This is not encryption
on
DNA Encryption
·
· Score: 1
For some reason this made me think of chaffing and winnowing, in which a series of messages are sent, each with a message authentication code (cryptographic checksum) that marks the message as authentic (i.e. you can check that it was signed with the sender's key). Any messages with fake MACs are noise that is filtered out by the receiver. It was described on the Net, and you can find a description under rsa.com somewhere.
The proposed scheme is really steganography, but if you encoded a number of messages into DNA, you would just need to attach MACs to the real messages and fake MACs to the fake messages, adding significant security.
I guess the marker that lets you pick out the signal from the mass of human DNA is rather like computing the MAC and finding that it matches the included MAC of a message.
Of course, you could just encrypt the signal messages instead - chaffing and winnowing is really intended for when you want to 'encrypt' without using any encryption, so to speak (only authentication is used to achieve confidentiality here.) However chaffing and winnowing helps by making the actual signal messages very hard to detect.
I don't claim to be expert, but something like 90% of all human DNA is 'junk' (now called something else I think), i.e. not known to code any proteins. This so-called junk DNA may in fact have some subtle purpose, but as long as the 'watermark' was coded into junk DNA it should be fine.
Re:Risky. Data just might code a deadly gene.
on
DNA Encryption
·
· Score: 1
I think the human gut is pretty good at handling random foreign DNA - after all, that's exactly what it does with the animal and/or vegetable DNA that you ingest every day... Viruses usually come protected by a protein sheath that enables them to penetrate cell walls and (if they are retroviruses) get into the nucleus. My cell biology's a bit rusty but I think this is right:)
Re:No need for watermarking people
on
DNA Encryption
·
· Score: 1
If you really want to categorise people like this, why not just have a big database with everyone's DNA fingerprint in it, with an attribute on each object that says what category it's in?
Despite the inaccuracies and comparing Linux to 3 Windows OSs including the unreleased Win2K, I thought the review was OK overall, and generally fair.
However, I don't see why NT can't be a server for multicast apps, in fact there is a Cisco IP/TV (video multicast) demo on my desk that runs on NT4 clients and servers. Multicast routing is not necessary for multicast server apps to run, IMO, though it would be necessary for a server with more than one NIC perhaps.
Agreed - I found that Red Hat 5.2 just recognised my PCMCIA modem/network card (from Ositech) on a Thinkpad 755 and the installer even let me NFS-install over the network using this card!
See http://www.winternals.com - they sell the FAT32 driver - the read-only version is free of charge, the r/w version is chargeable but not too expensive. I used it when I was dual booting Linux and NT, now I just use VMware (http://www.vmware.com) to run NT on top of Linux.
> Of course you're gonna lose out on stability and > scalability with NT, but frankly, you still > can't run alot of wintel type apps on linux > _today_.
Not true - just use VMware (www.vmware.com) and you can run Windows NT or Windows 98 (or even Win3.1) inside a virtual machine running on Linux. I do this all the time for a Windows-only email package, and it works fine - not suitable for games players, but great for running Wintel business/office apps on Linux.
The supported version is about $300 but I have the unsupported version at $99. Not cheap unfortunately but it does work very well, and it may be cheaper than junking Windows apps before they are obsolete.
Agreed, I've already played with Debian in a VM, and found out that Caldera 2.2 is allergic to being in a VM (at least for me).
I think it would be very smart to bundle the demo version of VMware on a commercial applications CD, such as the one in the Red Hat boxed set. Even better if they bundle the full one in a boxed set but then it's commercialware not freeware...
VMware is a real killer app for Linux - it's so convenient to have access to Windows stuff while apps vendors slowly make the migration to Linux.
Horses for courses - if you want to convert a Windows user to Linux, you are going to fail big-time if you hand them a Debian or Slackware CD, but they'll probably be fine with Caldera, Red Hat or (by all accounts) Mandrake. I've installed all the above except Slackware, and I'm quite happy with the character-based Debian 2.0 install, but I've got a reasonable background in various Unices.
Is it better that the only people using Linux are 'traditional Unix' types, or that a vast number of new users start using Linux, even if they expect and demand GUIs for everything?
Nobody is going to stop you using Debian or Slackware, and they're ideal for certain types of Linux usage. Just don't demand that everyone conforms to the way you view Linux...
You are both right... It would make most sense if/opt was a symlink to/usr/opt - still conformant to the Linux FHS but mapping onto the/usr partition to keep root small. Of course, if/opt is a separate partition the symlink should not be created.
Personally I have a single huge root partition that has everything, but I may live to regret this:)
As you point out, they are probably patenting some clever technique for virtualizing the apparently unvirtualizable, and this is most likely going to be patentable IMO.
I'd hope that they would not sue any truly open-source patent infringer, who is not making money off the back of the open source code.
HP's OpenView division is porting a number of network management products to Linux - Firehunter is already there and others are coming.
WatchGuard is a neat firewall device whose key feature is a team of security people who remotely update the rules and software as new exploits are discovered - all Linux based.
Interestingly, Cisco is going to bundle Apache with its NT-based web-based network management tools, apparently because it's too hard to set up the various flavours of MS IIS correctly.
Product support will follow Linux market share - the queso OS surveys make it quite clear that Linux is the dominant Web host with 30+ per cent market share, followed by Windows then Solaris - so it's likely that management tools focusing on web hosts will migrate to Linux quite quickly.
On a personal note... if there's anyone out there who wants bandwidth management / QoS support for Linux web hosting, please email me at rdonkin@orchestream.com as well as commenting here - I'm curious as to how much demand is out there, ideally from large enterprises or service providers. (My company, Orchestream, makes policy-based network management tools for QoS/CoS, and the Linux-DiffServ stuff looks like a very capable platform for this - I think web hosts will be the primary market for any Linux version.)
The key, as always, is to find the right person in a company (typically a product manager or marketing manager) and contact them with your request for Linux support, including details of the number of users, number of Linux hosts, so they can start salivating over the order size:)
I have to admit, this had me going until near the end - nicely done, but the very last paragraph gives the game away (emphasis added):
"The South to the Future World Wide Wire Service is a weekly feed of technology and media news commentary and *satire* published by the San Francisco Bay Guardian. Quotations attributed to public figures who are *satirized* are often true, but *sometimes invented*. Some fictional statements may, in fact, be true. Any other use of real names is accidental and coincidental."
So... While Pilots may help to give you RSI, or exacerbate existing RSI, they definitely don't give you a debilitating neurological disorder (other than the compelling need to manage your entire life on them, of course...).
New human interface technologies frequently have very unexpected health impacts - for example, keyboards and mice are a big factor in RSI, and immersive VR can give you motion sickness as well as more serious perceptual problems for some time afterwards. For details on the sorts of problems people get with RSI, see http://www.tifaq.com.
So I wouldn't rush to test this out - I'd hope that this would be tested for ergonomic and health issues, but no doubt it will be put on the market and then a few years later any health problems will surface.
It sounds entertaining but inherently problematic - what if they discover people's retinas aren't so tolerant to this sort of continual illumination? What if your retina gets a pattern 'burnt in' like the old Mac toolbar on the monitor - will you need a retina saver program to prevent this?? I know someone who may end up blind as a result of macular degeneration (a disease of the retina) so perhaps I'm biased - for more info on this see http://www.maculardegeneration.org/.
Excuse my cynicism, but I had RSI as a result of the last wave of new HCI technology, so I'll wait this one out.
Maltron layout and reduced hand motion
on
The Myth of QWERTY
·
· Score: 1
If you're interested in reducing hand motion, have a look at http://www.maltron.com/ - the Maltron keyboard has both an improved key layout and a physically sculpted two-dish design (hard to describe but also reduces hand movement by moving keys closer to fingertips at rest).
The Maltron layout was designed based on a corpus of English text and is claimed to greatly reduce hand motion. I use a Maltron keyboard with Qwerty layout but even that is a huge improvement over flat keyboards.
For anyone interested in RSI (and you should be, it's not uncommon, painful to treat, and easily prevented), there are some links at http://www.bigfoot.com/~rdonkin/
Just off the top of my head, here are some reasons to run VMware etc:
- cost - much cheaper to add $300 VMware and extra RAM than buy a new machine - any decent machine costs more than $450
- flexibility - run bleeding edge Linux/BSD/other kernels in a VM, hosted by production Linux installation. Run more than one guest OS at once.
- power usage and heat - run several OSs on one box without increasing heat dissipation - reduce electricity usage and help avoid global warming (seriously - PCs use a big chunk of the US electricity production!)
- tech support - boot a VM that is an exact replica of the one the user is running, without disturbing your normal session.
- Windows development - run a Windows NT guest OS to do email, another to do development (will sometimes crash but who cares), and a Win98 one and Win95 one for testing. If you hate Windows, remember that Linux will be the most stable host OS for VMware, most likely, and you are thereby introducing Linux onto a Windows developer's desktop - can't be all bad. And remember that this will support a Linux guest OS for when they decide they must port to Linux:)
- desk space - many of our developers have two PCs, one Windows and one Linux, which is a waste of desk space.
- testing small networks on single machine - providing you have enough RAM, you could run 5 or ten VMs at once, enough to test a small network of systems - ideal for testing client/server setups using different OSs without consuming a lot of hardware. My company makes network management software, so if some of the VMs could boot Linux router code we could test a complete network on a laptop. Handy for when our testers want to work from home rather than suffer up to 1.5 hours commute each way in London.
- research - develop and test new network protocols, or Beowulf apps, on a single machine
- Year 2000 testing - run a guest OS with its clock turned forward
- workload partitioning on large SMP systems - run different major apps, e.g. SAP R/3 and Oracle Applications, in different VMs with (hopefully) little impact of one on the other. Standard practice in mainframes and copied using hardware by Sun, HP, Amdahl, etc. Unix has traditionally been quite poor at workload partitioning - would be better to see it in the OS, but this is a useful stopgap. Very handy for server consolidation, where many small servers get merged into one or two huge servers.
- demonstrations - would be useful to demo distributed client/server type software, including showing how the system recovers from a (guest) OS/hardware crash, simulated by halting the VM.
It's worth remembering that the IT world is more complex than your own particular environment. Also, if Linux+VMware can be used to 'surround' Windows environments by running them in a more flexible way, it's only a short step to a more Linux-based environment.
VM/370, the IBM mainframe equivalent of VMware, was once about to be canned (it was an unofficial project done in R&D labs, a bit like Unix at AT&T), when the suits discovered that it was being used by the MVS team (MVS being the main operating system then and now for IBM mainframes) for development and testing, running several MVS instances on a single mainframe. Once they realised that MVS, although notionally a competitor to VM, was in fact depending on it, they kept the VM project going, and it's now available as VM/390.
As always, generalisations are dangerous - no doubt many hand held users use them just as appliances, but quite a few load them up with lots of shareware and write their own small apps (as I have done in Pocket C).
It's a question of whether you need the power of Linux on a handheld or not - if you develop on Linux, go for it, but if you just need address/calendar/notes etc, something like a Pilot or a Psion is cheaper and gets the job done quicker, compared to a more general purpose platform like Linux.
I've got nothing against Linux but I don't have enough of a desire to carry it in my shirt pocket, though of course it does run on a PalmPilot and I now have a 4MB Palm IIIx, so I could do... really I need VMware for the Palm platform so I can hypervise PalmOS and Linux, or at least run xcopilot under Linux on the Palm...
Actually I've seen whisky vending machines on the streets in Tokyo. Don't think this would work in England somehow with our underage drinking culture, but maybe the Japanese kids are better behaved?
The Japanese also have machines that sell hot coffee in cans, which double as handwarmers in the winter...
I'm now on my second Palm device (just upgraded to Palm IIIx, which has 4 MB and a better screen, and a free slot for future upgrades).
PalmOS has its limitations but the key thing is that the built-in apps are small, efficient, easy to learn, while there is a plethora of excellent third party apps, and PalmOS is very reliable as long as you run decent apps. Its uptime is measured in days, while my NT laptop (with 96MB) has an uptime of days before a reboot is required for some reason.
Above all, Palm devices are really quick - I can search 320 address book entries in 4 seconds on the IIIx, and in 10 seconds on my P233 laptop using Outlook and NT (identical data and search). Outlook is even slower searching our company contacts database, and I can probably now fit this into my IIIx if needed.
And of course, there are Palm development tools available for almost every operating system - Windows, Mac, Linux, *nix, etc. Try syncing a WinCE machine with non-Microsoft apps and OSs...
I was at the IETF last week, and it's interesting to note that at least some of the Microsoft people there use Palm devices!
Have a look at their website, http://www.vmware.com/ - it's pretty informative, with architecture diagrams etc. Looks like a VM/390 style virtual machine hypervisor, i.e. each guest OS is transparently multitasked with other OSs. The only reason there are Linux and NT versions, IMO, is so that the control applications to toggle between OSs etc can run under an OS, and are therefore a bit easier to access. Any version should be able to run a wide range of guest OSs including Linux, NT, FreeBSD, Win98, etc.
I've put my name down for the beta - it's pricey but then the market is probably quite limited, and it's not trivial to develop.
Slashdot Mirror
Unfortunately the default permissions on NT are typically wide open, e.g. anyone can write to system directory and many registry keys. There was a time when Unix was like this, maybe Win2000 will fix this.
For some reason this made me think of chaffing and winnowing, in which a series of messages are sent, each with a message authentication code (cryptographic checksum) that marks the message as authentic (i.e. you can check that it was signed with the sender's key). Any messages with fake MACs are noise that is filtered out by the receiver. It was described on the Net, and you can find a description under rsa.com somewhere.
The proposed scheme is really steganography, but if you encoded a number of messages into DNA, you would just need to attach MACs to the real messages and fake MACs to the fake messages, adding significant security.
I guess the marker that lets you pick out the signal from the mass of human DNA is rather like computing the MAC and finding that it matches the included MAC of a message.
Of course, you could just encrypt the signal messages instead - chaffing and winnowing is really intended for when you want to 'encrypt' without using any encryption, so to speak (only authentication is used to achieve confidentiality here.) However chaffing and winnowing helps by making the actual signal messages very hard to detect.
I don't claim to be expert, but something like 90% of all human DNA is 'junk' (now called something else I think), i.e. not known to code any proteins. This so-called junk DNA may in fact have some subtle purpose, but as long as the 'watermark' was coded into junk DNA it should be fine.
I think the human gut is pretty good at handling random foreign DNA - after all, that's exactly what it does with the animal and/or vegetable DNA that you ingest every day... Viruses usually come protected by a protein sheath that enables them to penetrate cell walls and (if they are retroviruses) get into the nucleus. My cell biology's a bit rusty but I think this is right :)
If you really want to categorise people like this, why not just have a big database with everyone's DNA fingerprint in it, with an attribute on each object that says what category it's in?
True about NT4 and FAT32, but Win2000 will support FAT32.
Despite the inaccuracies and comparing Linux to 3 Windows OSs including the unreleased Win2K, I thought the review was OK overall, and generally fair.
However, I don't see why NT can't be a server for multicast apps, in fact there is a Cisco IP/TV (video multicast) demo on my desk that runs on NT4 clients and servers. Multicast routing is not necessary for multicast server apps to run, IMO, though it would be necessary for a server with more than one NIC perhaps.
Agreed - I found that Red Hat 5.2 just recognised my PCMCIA modem/network card (from Ositech) on a Thinkpad 755 and the installer even let me NFS-install over the network using this card!
See http://www.winternals.com - they sell the FAT32 driver - the read-only version is free of charge, the r/w version is chargeable but not too expensive. I used it when I was dual booting Linux and NT, now I just use VMware (http://www.vmware.com) to run NT on top of Linux.
Windows 2000 supports FAT32 r/w.
> Of course you're gonna lose out on stability and
> scalability with NT, but frankly, you still
> can't run alot of wintel type apps on linux
> _today_.
Not true - just use VMware (www.vmware.com) and you can run Windows NT or Windows 98 (or even Win3.1) inside a virtual machine running on Linux. I do this all the time for a Windows-only email package, and it works fine - not suitable for games players, but great for running Wintel business/office apps on Linux.
The supported version is about $300 but I have the unsupported version at $99. Not cheap unfortunately but it does work very well, and it may be cheaper than junking Windows apps before they are obsolete.
Agreed, I've already played with Debian in a VM, and found out that Caldera 2.2 is allergic to being in a VM (at least for me).
I think it would be very smart to bundle the demo version of VMware on a commercial applications CD, such as the one in the Red Hat boxed set. Even better if they bundle the full one in a boxed set but then it's commercialware not freeware...
VMware is a real killer app for Linux - it's so convenient to have access to Windows stuff while apps vendors slowly make the migration to Linux.
Horses for courses - if you want to convert a Windows user to Linux, you are going to fail big-time if you hand them a Debian or Slackware CD, but they'll probably be fine with Caldera, Red Hat or (by all accounts) Mandrake. I've installed all the above except Slackware, and I'm quite happy with the character-based Debian 2.0 install, but I've got a reasonable background in various Unices.
Is it better that the only people using Linux are 'traditional Unix' types, or that a vast number of new users start using Linux, even if they expect and demand GUIs for everything?
Nobody is going to stop you using Debian or Slackware, and they're ideal for certain types of Linux usage. Just don't demand that everyone conforms to the way you view Linux...
You are both right... It would make most sense if /opt was a symlink to /usr/opt - still conformant to the Linux FHS but mapping onto the /usr partition to keep root small. Of course, if /opt is a separate partition the symlink should not be created.
:)
Personally I have a single huge root partition that has everything, but I may live to regret this
As you point out, they are probably patenting some clever technique for virtualizing the apparently unvirtualizable, and this is most likely going to be patentable IMO.
I'd hope that they would not sue any truly open-source patent infringer, who is not making money off the back of the open source code.
HP's OpenView division is porting a number of network management products to Linux - Firehunter is already there and others are coming.
:)
WatchGuard is a neat firewall device whose key feature is a team of security people who remotely update the rules and software as new exploits are discovered - all Linux based.
Interestingly, Cisco is going to bundle Apache with its NT-based web-based network management tools, apparently because it's too hard to set up the various flavours of MS IIS correctly.
Product support will follow Linux market share - the queso OS surveys make it quite clear that Linux is the dominant Web host with 30+ per cent market share, followed by Windows then Solaris - so it's likely that management tools focusing on web hosts will migrate to Linux quite quickly.
On a personal note... if there's anyone out there who wants bandwidth management / QoS support for Linux web hosting, please email me at rdonkin@orchestream.com as well as commenting here - I'm curious as to how much demand is out there, ideally from large enterprises or service providers. (My company, Orchestream, makes policy-based network management tools for QoS/CoS, and the Linux-DiffServ stuff looks like a very capable platform for this - I think web hosts will be the primary market for any Linux version.)
The key, as always, is to find the right person in a company (typically a product manager or marketing manager) and contact them with your request for Linux support, including details of the number of users, number of Linux hosts, so they can start salivating over the order size
I have to admit, this had me going until near the end - nicely done, but the very last paragraph gives the game away (emphasis added):
...).
"The South to the Future World Wide Wire Service is a weekly feed of technology and media news commentary and *satire* published by the San Francisco Bay Guardian. Quotations attributed to public figures who are *satirized* are often true, but *sometimes invented*. Some fictional statements may, in fact, be true. Any other use of real names is accidental and coincidental."
So... While Pilots may help to give you RSI, or exacerbate existing RSI, they definitely don't give you a debilitating neurological disorder (other than the compelling need to manage your entire life on them, of course
New human interface technologies frequently have very unexpected health impacts - for example, keyboards and mice are a big factor in RSI, and immersive VR can give you motion sickness as well as more serious perceptual problems for some time afterwards. For details on the sorts of problems people get with RSI, see http://www.tifaq.com.
So I wouldn't rush to test this out - I'd hope that this would be tested for ergonomic and health issues, but no doubt it will be put on the market and then a few years later any health problems will surface.
It sounds entertaining but inherently problematic - what if they discover people's retinas aren't so tolerant to this sort of continual illumination? What if your retina gets a pattern 'burnt in' like the old Mac toolbar on the monitor - will you need a retina saver program to prevent this?? I know someone who may end up blind as a result of macular degeneration (a disease of the retina) so perhaps I'm biased - for more info on this see http://www.maculardegeneration.org/.
Excuse my cynicism, but I had RSI as a result of the last wave of new HCI technology, so I'll wait this one out.
If you're interested in reducing hand motion, have a look at http://www.maltron.com/ - the Maltron keyboard has both an improved key layout and a physically sculpted two-dish design (hard to describe but also reduces hand movement by moving keys closer to fingertips at rest).
The Maltron layout was designed based on a corpus of English text and is claimed to greatly reduce hand motion. I use a Maltron keyboard with Qwerty layout but even that is a huge improvement over flat keyboards.
For anyone interested in RSI (and you should be, it's not uncommon, painful to treat, and easily prevented), there are some links at http://www.bigfoot.com/~rdonkin/
Just off the top of my head, here are some reasons to run VMware etc:
:)
- cost - much cheaper to add $300 VMware and extra RAM than buy a new machine - any decent machine costs more than $450
- flexibility - run bleeding edge Linux/BSD/other kernels in a VM, hosted by production Linux installation. Run more than one guest OS at once.
- power usage and heat - run several OSs on one box without increasing heat dissipation - reduce electricity usage and help avoid global warming (seriously - PCs use a big chunk of the US electricity production!)
- tech support - boot a VM that is an exact replica of the one the user is running, without disturbing your normal session.
- Windows development - run a Windows NT guest OS to do email, another to do development (will sometimes crash but who cares), and a Win98 one and Win95 one for testing. If you hate Windows, remember that Linux will be the most stable host OS for VMware, most likely, and you are thereby introducing Linux onto a Windows developer's desktop - can't be all bad. And remember that this will support a Linux guest OS for when they decide they must port to Linux
- desk space - many of our developers have two PCs, one Windows and one Linux, which is a waste of desk space.
- testing small networks on single machine - providing you have enough RAM, you could run 5 or ten VMs at once, enough to test a small network of systems - ideal for testing client/server setups using different OSs without consuming a lot of hardware. My company makes network management software, so if some of the VMs could boot Linux router code we could test a complete network on a laptop. Handy for when our testers want to work from home rather than suffer up to 1.5 hours commute each way in London.
- research - develop and test new network protocols, or Beowulf apps, on a single machine
- Year 2000 testing - run a guest OS with its clock turned forward
- workload partitioning on large SMP systems - run different major apps, e.g. SAP R/3 and Oracle Applications, in different VMs with (hopefully) little impact of one on the other. Standard practice in mainframes and copied using hardware by Sun, HP, Amdahl, etc. Unix has traditionally been quite poor at workload partitioning - would be better to see it in the OS, but this is a useful stopgap. Very handy for server consolidation, where many small servers get merged into one or two huge servers.
- demonstrations - would be useful to demo distributed client/server type software, including showing how the system recovers from a (guest) OS/hardware crash, simulated by halting the VM.
It's worth remembering that the IT world is more complex than your own particular environment. Also, if Linux+VMware can be used to 'surround' Windows environments by running them in a more flexible way, it's only a short step to a more Linux-based environment.
VM/370, the IBM mainframe equivalent of VMware, was once about to be canned (it was an unofficial project done in R&D labs, a bit like Unix at AT&T), when the suits discovered that it was being used by the MVS team (MVS being the main operating system then and now for IBM mainframes) for development and testing, running several MVS instances on a single mainframe. Once they realised that MVS, although notionally a competitor to VM, was in fact depending on it, they kept the VM project going, and it's now available as VM/390.
I don't believe it - now we are going to have politically correct computer jargon??? Please tell me you're kidding...
Linux already runs on the PalmPilot with extra memory (2MB total I think is needed), but it's still quite bleeding edge.
As always, generalisations are dangerous - no doubt many hand held users use them just as appliances, but quite a few load them up with lots of shareware and write their own small apps (as I have done in Pocket C).
... really I need VMware for the Palm platform so I can hypervise PalmOS and Linux, or at least run xcopilot under Linux on the Palm...
It's a question of whether you need the power of Linux on a handheld or not - if you develop on Linux, go for it, but if you just need address/calendar/notes etc, something like a Pilot or a Psion is cheaper and gets the job done quicker, compared to a more general purpose platform like Linux.
I've got nothing against Linux but I don't have enough of a desire to carry it in my shirt pocket, though of course it does run on a PalmPilot and I now have a 4MB Palm IIIx, so I could do
Actually I've seen whisky vending machines on the streets in Tokyo. Don't think this would work in England somehow with our underage drinking culture, but maybe the Japanese kids are better behaved?
The Japanese also have machines that sell hot coffee in cans, which double as handwarmers in the winter...
I'm now on my second Palm device (just upgraded to Palm IIIx, which has 4 MB and a better screen, and a free slot for future upgrades).
PalmOS has its limitations but the key thing is that the built-in apps are small, efficient, easy to learn, while there is a plethora of excellent third party apps, and PalmOS is very reliable as long as you run decent apps. Its uptime is measured in days, while my NT laptop (with 96MB) has an uptime of days before a reboot is required for some reason.
Above all, Palm devices are really quick - I can search 320 address book entries in 4 seconds on the IIIx, and in 10 seconds on my P233 laptop using Outlook and NT (identical data and search). Outlook is even slower searching our company contacts database, and I can probably now fit this into my IIIx if needed.
And of course, there are Palm development tools available for almost every operating system - Windows, Mac, Linux, *nix, etc. Try syncing a WinCE machine with non-Microsoft apps and OSs...
I was at the IETF last week, and it's interesting to note that at least some of the Microsoft people there use Palm devices!
Have a look at their website, http://www.vmware.com/ - it's pretty informative, with architecture diagrams etc. Looks like a VM/390 style virtual machine hypervisor, i.e. each guest OS is transparently multitasked with other OSs. The only reason there are Linux and NT versions, IMO, is so that the control applications to toggle between OSs etc can run under an OS, and are therefore a bit easier to access. Any version should be able to run a wide range of guest OSs including Linux, NT, FreeBSD, Win98, etc.
I've put my name down for the beta - it's pricey but then the market is probably quite limited, and it's not trivial to develop.